Intensity Modulated Radiation Therapy (IMRT) seeks to deliver highly conformal tumorcidal doses to selected target volumes while conformably avoiding nearby normal tissues and critical structures. IMRT assumes that the optimized planned doses can be precisely and accurately delivered to the patient and that the final plan is optimal relative to the desired treatment objectives. Yet, current IMRT planning systems use simplified fast dose calculations during plan optimization and divides optimization and plan delivery into separate processes. The simplified dose computations can result in dose prediction errors (DPEs), which are differences between actual and computed doses, of 10% or more, while optimization with algorithms that have DPEs and the transformation from optimized to deliverable parameter spaces lead to optimization convergence errors (OCEs). OCEs prevent the true optimal plan from being found and can result in critical structure doses that are as much as 20% greater than necessary. Since dose differences of 3%-5% can produce clinically detectable changes in response, DPEs and OCEs may significantly impact clinical outcome. The goals of this project are: (1) To develop accurate IMRT dose-calculation methods that reduce DPEs to clinically insignificant levels. Monte Carlo (MC) methods will be used to evaluate the sources and clinical impacts of DPEs caused by patient heterogeneities, incident fluence prediction, and patient set-up errors and to identify and develop methods to reduce IMRT dose-calculation errors to less than 2% for all patient cases. (2) To develop optimization processes that reduces OCEs to clinically insignificant levels. This will be achieved by incorporating algorithms, identified in goal 1, that accurately model radiation transport through the multi-leaf collimator and the patient into the IMRT optimization process. Such deliverable optimization will allow each intensity-modulated beam to partially compensate for the limitations of other beams and may result in significantly reduced doses to critical Istructures for the same planning target volume dose. (3) To improve the computational efficiency of accurate Ideliverable-optimization processes developed in goal 2 to make them clinically practical without compromising plan laccuracy or optimality. This will allow accurate, deliverable optimized plans to be used for routine clinical IMRT, Iwhere previously it was not practical due to the excessive calculation time required. The long-term objectives of' this project are to develop rapid, accurate IMRT dose calculation and optimization} methods that result in minimal dose prediction and optimization convergence errors, to use these more accurate IRMT to accumulate more reliable dose-response data, and to improve patient outcomes through more conformal.